Environmental performances’ impact on financial
performance: Do multinational companies suffer financially
more than domestic corporations?
Michel Rötter – S3479129
MSc International Financial Management
University of Groningen
Supervisor: Prof. Dr. L.J.R. Scholtens
Co-assessor: Dr. Ambrogio Dalò
7
thJune 2019
Abstract
1. Introduction
Based on the rising threat of climate change and global warming for civilisations across the globe, corporations are experiencing a greater demand for environmentally conscious actions, particularly in developed nations (Iwata and Okada, 2011). This demand arises from various stakeholders, including regulatory bodies through implemented laws, non-governmental organizations, local communities and financial agencies (Martin et al., 2015; Iwata and Okada, 2011). As a consequence, the financial performance of companies can be affected. Corporations may face fines and penalties in case of non-compliance with stakeholder demands and additional destruction of reputational value or, more drastically, a boycott of products. Ultimately, failure to comply to standards can increase a company’s risk premium, which can result in avoidance from future investors or customers (Preston and O’Bannon, 1997; Iwata and Okada, 2011). Proactive environmental management practices have been found to be essential parts of corporations’ business operations, as these can lead to an increase in positive reputation and improvement of operational efficiency (Rondinelli and Berry, 2000; Freeman and Evan, 1990; Hill and Jones, 1992).
study is if the assumed negative effect from environmental on financial performance is stronger for MNCs in comparison to domestic companies (DCs). For the purpose of answering this question, environmental performance, which is measured in this study as pollutant emissions to net sales of a company, is interacted with a proxy for MNCs. Then, the respective environmental performance measure together with the interaction term is regressed on one of five financial performance variables.
The international domain in this thesis is covered through the analysis of the interaction effect of MNCs in comparison to DCs and through a broad multinational sample of corporations covering more than 30 countries. Furthermore, analysing the changes in financial performance through both accounting- and market-based figures, as well as using proxies for risk, addresses the financial dimension. At last, management implications can be derived based on the results, as it is in the management’s best interest to maximize shareholder value while simultaneously complying with rules and norms imposed by the relevant jurisdictions and societies.
respective effects on financial performance. Additionally, I contribute to prior research in this field with the investigation if MNCs suffer financially more than domestic firms from an increase in pollutant emissions.
The main results of the research conducted in this paper are that for this data sample, a bi-directional causality between environmental and financial performance is present. This means that prior environmental performance affects financial performance due to a firm’s external reputation. The other causal direction is also found, where financial performance precedes environmental performance based on available resource that can be allocated to reducing emissions. A sensitivity analysis employed with more lags broadly confirms the outcome of the Granger-causality tests. Moreover, in the majority of cases, an increase in pollutant emissions and thus an increase in environmental performance results in a decrease in financial performance. This effect is generally stronger for the case of an MNC in comparison to DCs, which means that multinational corporations suffer financially more from an increase in emissions than domestic firms. This research has important managerial implications. It shows that when managers ignore pollutant emission levels of a company, detrimental effects on the financial performance of the corporation can follow. This is specifically applicable for managers in MNCs, as the negative effect on financial performance of increased pollutant emissions is stronger for multinational corporations than for domestic companies.
The remainder of this paper proceeds as follows. First, the theoretical background is introduced, supported by empirical results from various studies. Then, data and methods used in this paper are reported. Next, the results of the causality tests as well as regression analyses are presented. Finally, I will discuss the conclusions and address limitations of this research.
2. Literature and hypotheses
hypothesis. Afterwards, an evaluation of the literature on the effect of environmental performance on financial performance is presented. This is followed by defining the second hypothesis. Subsequently, literature addressing the differences between MNCs and DCs characteristics is examined. Specifically, the stakeholder theory is used to differentiate them. Then, the last hypothesis is verbalized.
The relationship between environmental performance and financial performance has been subject to various research studies that offer insight into the underlying theory. In general, Baron (2000) stresses that a corporation’s performance is affected by its strategies and operations that comprise both its market and non-market environment. Environmental performance can, in this sense, be allocated to the non-market environment. According to Preston and O’Bannon (1997), corporate social performance (CSP) comprises the environmental performance of a company, the quality of its products and services and the ability to hire and develop superior employees. Environmental performance is the only component of CSP that can measure pollutant emissions and capture the environmental accountability of a corporation.
Table 1: Overview of different theories present in studies concerning the causal relationship between environmental and financial performance.
Causal direction Theoretical reasoning Study
EP ® FP Stakeholder theory Preston and O’Bannon (1997) Freeman (2010)
Cornell and Shapiro (1987) Trade-off theory Preston and O’Bannon (1997)
Friedman (1970) Palmer et al. (1995) FP ® EP Slack resource theory Orlitzky et al. (2003)
Russo and Fouts (1997) McGuire et al. (1988)
Managerial opportunism theory Allouche and Laroche (2005) Preston and O’Bannon (1997)
EP « FP Synergy Preston and O’Bannon (1997)
causal relationship between CSP and CFP is possible as CSP and CFP can interact in a synergetic relationship. A positive synergy can lead to a virtuous circle, in which superior CSP leads to greater CFP, which then in turn leads to an even further increase in CSP (Waddock and Graves, 1997). However, a negative synergy in form of a vicious circle may also exist (Preston and O’Bannon, 1997).
Scholtens (2008) analysed the direction of causality empirically for a variety of different CSP variables. Two different methods, distributed lags and Granger causation, are used to arrive at a conclusion about a causal direction. Scholtens (2008) finds that financial performance generally precedes social performance more often than vice-versa. Nakao et al. (2007) analyse statistical causality between environmental and financial performance variables. For this, they also utilize the Granger causality test methodology. Nakao et al. (2007) find that there is no unidirectional Granger causality for either direction between environmental performance and financial performance when a time-window of four years is chosen. A statistically significant bi-directional causality at the 1% level can be found in a time-frame of two years (Nakao et al., 2007). In this study, environmental performance, as one of CSP’s components, is used since it addresses pollutant emissions and the firm’s behaviour in dealing with global warming. To investigate a causal relationship between environmental and financial performance in this sample, the first hypothesis is formulated as follows:
H1: Environmental (financial) performance does Granger-cause financial (environmental) performance
The corresponding null hypothesis is that there is no Granger-causation between the respective variables.
environmental to financial performance (see, e.g. Hibiki and Managi, 2010; Mahoney and Roberts, 2007; Wagner et al., 2002) based on the stakeholder theory and trade-off theory. Therefore, this research will equally focus on this orientation. Environmental performance can positively or negatively affect financial performance. In line with the stakeholder theory mentioned in table 1, Iwata and Okada (2011) claim that different stakeholders, such as local communities, consumers and investors, are conscious about corporate environmental management. Every stakeholder may have different interests in the environmental actions of a company, yet they are united in scrutinizing a company if it is unsuccessful in adhering to environmental standards. When a company fails to meet expectations of stakeholders (e.g. reducing pollutant emissions) market fears will arise. These market fears can then lead to a rise in the risk premium and ultimately to a loss in profit contingencies (Cornell and Shapiro, 1987). Further, if a corporation disregards environmental policy or experiences externally invoked events, such as an environmental disaster, the company may suffer through penalties. Moreover, a loss of reputation can occur, which may lead to a boycott of products (Iwata and Okada, 2011). On the other side, corporations with superior environmental performance may attract highly qualified employees, reduce costs and increase operational efficiency, which will eventually lead to an increase in market opportunities and an improvement in the relationship with stakeholders (Gallego‐Álvarez et al., 2014; Porter and van der Linde, 1995). A superior financial performance can thus be achieved through the simultaneous coordination and prioritization of different stakeholder interests (Freeman and Evan, 1990; Hill and Jones, 1992).
resources may be drawn away from the firm. Allouche and Laroche (2005) confirm that increasing attention on emission reduction leaves a company in relative disadvantage compared to other corporations who do not focus that proactively on environmental performance. Now, after providing the theoretical background for the relationship, I analyse the empirical studies on this topic. Table 2 presents empirical studies that investigated the effect of an increase from environmental performance on financial performance.
Table 2: Overview of empirical studies that analyse the effect of environmental performance on financial performance. Effect of EP on FP Environmental performance proxy Financial performance proxy Study Positive effect
Environmental score ROA Russo and Fouts (1997) ROA, ROE, Tobin’s Q-1,
EPS Nakao et al. (2007)
Tobin’s Q, Excess return ROE, Business Risk, Beta
Gonenc and Scholtens (2017)
Environmental index ROCE, ROS, ROE Wagner et al. (2002)
CO2 productivity Tobin’s Q Nishitani and Kokubu (2012)
Toxic waste recycled / Total toxic waste
Annual stock returns Al-Tuwaijri et al. (2004)
Emission reduction ROA, ROE, ROS Hart and Ahuja (1996) Negative
effect
Toxic chemical emission / revenue; number of lawsuits
Tobin’s Q Konar and Cohen (2001)
Toxic chemical emission
Tobin’s Q King and Lennox (2001) Variation in CO2
emission
ROA, ROE Alvarez (2012) CO2 emissions / Net
sales, Waste
emission / Net sales
ROA, ROE, ROI, ROIC, Tobin’s Q-1, ln(Tobin’s Q)
Iwata and Okada (2011)
Environmental index ROA, ROE Horváthová (2012)
Water pollution Market value of equity Cormier and Magnan (1997) Compliance index Monthly stock returns Filbeck and Gorman (2004)
there is no consensus among researchers on how to approximate environmental performance. Several studies use an environmental score as a proxy and find a positive effect of environmental performance on financial performance (Russo and Fouts, 1997; Nakao et al., 2007; Gonenc and Scholtens, 2017). This can be explained by the fact that higher environmental scores indicate initiatives to meet environmental demands from stakeholders, compliance records and activities to reduce emissions. In consequence, this supports the stakeholder theory and rejects the trade-off theory, as allocated funds towards environmentally-conscious actions do not seem to negatively affect the firm’s financials. If a firm reduces its pollutant emissions, it can also expect a positive effect on its financial performance, as shown by Nishitani and Kokubu (2012), Al-Tuwaijri et al. (2004) and Hart and Ahuja (1996). Further, researchers employ different proxies for financial performance. The two popular approaches are backward-looking accounting figures, that can be subject to manipulation, and forward-backward-looking market performance measures (Wang et al., 2014). Gonenc and Scholtens (2017) and Nakao et al. (2007) employ a mix of accounting figures, such as return on equity, and market-oriented indicators with Tobin’s Q and excess returns. The remaining mentioned studies in table 2 for a positive effect of environmental on financial performance apply a single focus on either financial performance proxy (Wagner et al., 2002; Russo and Fouts, 1997; Al-Tuwaijri et al., 2004; Hart and Ahuja 1996). The positive results of an improved environmental performance found by different studies apply to both accounting- and market-based financial performance proxies, as shown in table 2.
stakeholder theory and Cornell and Shapiro’s (1987) notion that companies will be scrutinized by stakeholders when emissions are increased, which then leads to losses in revenue streams and subsequent financial instability. While King and Lennox (2001) and Konar and Cohen (2001) use toxic chemical emissions, a proxy which includes different pollutants, Alvarez (2012) and Cormier and Magnan (1997) focus on a specific pollutant in their analyses. The latter approach can bring greater clarity by isolating the effects of a single pollutant, yet such an analysis may be biased through its narrow focus. Horváthová (2012) and Filbeck and Gorman (2004) use indices to investigate if the effect of environmental performance on financial performance is negative. Filbeck and Gorman (2004) find support for the trade-off hypothesis. When a company allocates investments towards meeting stakeholder demands, the compliance index increases. However, the effect on the financial performance is in this case negative. Therefore, investors perceive this as inadequate allocation of funds that would provide more financial benefit at the core practices of a company. In Horváthová’s (2012) case, the environmental index is created based on emission levels. Therefore, the results show that with an increase in emissions the index rises, which then negatively affects financial performance. This result from Horváthová (2012) is further evidence for the stakeholder theory. For this study, environmental performance is approximated in coherence with studies that observe a negative impact of increased environmental performance on financial performance. Specifically, emission levels in relation to net sales are used as environmental performance proxy. Therefore, an increase in emission and thus in environmental performance is hypothesized to have a negative effect on financial performance. This is grounded in the stakeholder theory illustrated before, which explains that failure to comply with stakeholder expectations results in adverse effects for a firm’s financial performance. The corresponding hypothesis is formulated as follows:
The corresponding null hypothesis is that there is no effect of environmental performance on financial performance.
In the next section of this chapter, I will analyse how MNCs can be affected differently than domestic corporations by an effect of an increase in environmental performance on financial performance. Table 3 presents studies that have used the stakeholder theory and applied it to MNCs in order to distinguish them from domestic companies.
Table 3: Overview of studies that address how the stakeholder theory and its characteristics have specific application to MNCs.
Theory Application to MNC Study
Stakeholder theory Greater visibility of actions Bouquet and Deutsch (2008) Albinger and Freeman (2000) Bansal and Roth (2000) Kolk and van Tulder (2003) Kostova and Zaheer (1999) Multicultural dimension
creates complexity Robertson and Crittenden (2003) Indifferent attitude towards
society Kennelly and Lewis (2003)
Liability of foreignness Chen et al. (2016)
of MNCs, which can improve the public image and legitimacy (Van Tulder and Kolk, 2001; Christmann, 2004; Bouquet and Deutsch, 2008; Albinger and Freeman, 2000; Bansal and Roth, 2000). Since MNCs act in dimensions of greater cultural diversity than DCs, they face complexity in addressing stakeholder demands, as certain practices may not be culturally or legally acceptable. In addition, MNCs can be indifferent or detrimental in their actions, as they may abuse their supranational power to shift standards and operations across borders in search for cost savings in countries with low regulatory standards and stakeholder demands (Sethi, 2003; Robertson and Crittenden, 2003; Kennelly and Lewis, 2003). When neglecting stakeholder interests, MNCs can face greater consequences than DCs because their geographical expansion can be hindered, and the liability of foreignness can be exacerbated. MNCs that are not acting in the stakeholder’s interest may suffer greater damage of reputation and lost profit contingencies than DCs due to the heightened visibility of their actions (Chen et al., 2016, Christmann and Taylor, 2002).
H2b: Multinational corporations experience stronger negative effects on financial performance from an increase of environmental performance
The relevant null hypothesis is that MNCs are not affected differently than DCs by a negative effect on financial performance from increased environmental performance.
3. Data and methodology
In this chapter, the basic models for statistical evaluation of the hypotheses are first presented. Then, the data sample and the respective descriptive statistics are shown. Subsequently, the methods to approximate the models are outlined.
At first, the causality issue between environmental and financial performance is addressed with the use of Granger causality tests. The following two equations show the respective Granger causality analysis model for evaluating hypothesis 1:
𝐹𝑃#$% = 𝛼 + 𝛽*𝐸𝑃#$%,*+ 𝛽-𝐹𝑃#$%,*+ 𝜀#% (1)
𝐸𝑃#$% = 𝛼 + 𝛽*𝐹𝑃#$%,*+ 𝛽-𝐸𝑃#$%,*+ 𝜗#% (2) where 𝑖 describes the firm, 𝑗 denotes the financial or environmental performance measure used, 𝑡 is a subscript for each year, 𝛼 is a constant, 𝛽*is the coefficient for the lagged independent variable, 𝛽- is the coefficient for the lagged dependent variable and 𝜀 as well as 𝜗 are the respective errors terms.
In order to test hypotheses 2a statistically, a multivariate regression model is constructed to investigate the effect of environmental performance on financial performance. The model specification is as follows:
𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒#$%= a + 𝛽* 𝑆𝑖𝑧𝑒#%+ 𝛽- 𝐶𝑎𝑝𝐼𝑛𝑡#% + 𝛽F 𝐺𝑟𝑜𝑤𝑡ℎ#%
where 𝑖 describes the firm, 𝑗 denotes the financial performance measure, 𝑡 is a subscript for each year and 𝜀 is the error term. Further, 𝑚 denotes the respective variable for environmental performance used in the model. In addressing hypothesis 2b, the presented regression model has to be modified, as the interaction term between environmental performance and multinationality is included. This leads to the following specification of the multivariate regression model for hypothesis 2:
𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒#$%= 𝑎 + 𝛽* 𝑆𝑖𝑧𝑒#%+ 𝛽- 𝐶𝑎𝑝𝐼𝑛𝑡#%+ 𝛽F 𝐺𝑟𝑜𝑤𝑡ℎ#% + 𝛽J 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒#% + 𝛽N 𝐸𝑃#O%+ 𝛽V𝜃𝑀𝑁𝐶#%+ 𝛽Z𝐸𝑃#O% 𝑥 𝜃𝑀𝑁𝐶#%+ 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐹𝐸 + 𝑌𝑒𝑎𝑟𝐹𝐸 + 𝜀#% (4)
where all subscripts are similar to the basic regression equation. In addition, 𝜃 signals the use of a dummy variable.
Companies from three countries, the United States, the United Kingdom and Japan, comprise more than 53% of the entire sample. To secure robustness of the results, a sensitivity analysis for the effect of environmental performance on financial performance is separately conducted for these three countries. The results from this evaluation are reported in the respective results section of this paper. Further, various industries are subject to investigation in this study whereas previous academic work mostly focused on a specific industry. Moreover, the time-period of ten years selected for this study is superior to previously conducted analyses because it is longer (see, e.g. Alvarez, 2012; Nakao et al., 2007).
In respect to equation 1 and 2, Scholtens (2008) and Nakao et al. (2007) utilize a comparable model to investigate Granger causality. Scholtens (2008) illustrates that Granger causality tests are in fact rather about precedence than causation in the traditional sense. Hence, Granger causality can be seen as a lead-lag analysis. Moreover, Scholtens (2008) argues that if environmental performance causes financial performance, lags of environmental performance should be significant in the equation for financial performance. This is in coherence with Horváthová (2012), as previous research further stresses that time elapses until actions of emission reductions materialize (see, e.g. Hart and Ahuja, 1996; Konar and Cohen, 2001). Concerning equation 3 and 4, it can be stressed that meta-analyses found that a multivariate regression model is the most commonly used technique to arrive at statistical results for the effect of environmental performance on financial performance (Orlitzky et al., 2003; Horváthová, 2012).
in the regression analyses. Further, the errors have to be uncorrelated with each other, hence no autocorrelation should be present. This is tested with a Wooldridge test for autocorrelation. The corresponding results in appendix B.2 show that there is evidence for autocorrelation in this sample. However, autocorrelation can also be treated with the use of robust standard errors in the regression analyses. At last, multicollinearity may arise if explanatory variables are highly correlated with each other. In reference to the correlation matrix in appendix B.3, one can see that multicollinearity is not present for the variables used in this study. To reduce possible endogeneity problems, I use fixed effects estimation for equation 3 and equation 4, as the sample consists of ten years of unbalanced panel data. With the intention to account for unobserved differences among the SIC-classified industries that may alter the estimation results, I employ industry-fixed effects. Additionally, year-fixed effects are used, which fulfil a similar purpose of minimizing unobserved effects caused between different years of observation. The inclusion of fixed effects reduces but does not completely remove the possibility that a relationship can be affected by omitted variable bias. This is in line with Lee et al. (2015) who claim that a fixed effects model helps to partly resolves endogeneity biases caused by omitted variables. The possible cost of reducing the omitted variable bias is that a substantial amount of signal in the data could be lost. This translates to a reduction in the power of the analysis. According to Lee et al. (2015), the fixed effects model can also be used to account for unobserved heterogeneity.
management systems as a proxy for environmental performance (Halkos and Sepetis, 2007). Another common approach is to construct environmental variables that reflect environmental performance on the firm level, as practiced by Hart and Ahuja (1996). Substantial variation in data availability between these variables leads to a low comparability between different proxies (Horváthová, 2012). In coherence with Iwata and Okada (2011) and Alvarez (2012), this study will use the total carbon dioxide (CO2) emissions by a corporation divided by net sales.
Furthermore, the emissions of NOx and SOx in relation to net sales are used to estimate environmental performance of a corporation, which is in coherence with Wagner et al. (2002). In addition, water pollutants and total waste as well as the energy consumption of each individual firm, all divided by the corresponding net sales, are used as proxies for environmental performance. These three proxies are conceptualized by Gonenc and Scholtens (2017), who create ratings based on various environmental dimension variables. All environmental performance variables are calculated so that an increase in emissions results in an increase of environmental performance. Therefore, it is important to stress that the term performance in this case does not refer to an improvement of the environmental practices of a company but rather the opposite, as more pollution is emitted in relation to the company’s net sales.
An alternative approach is to measure financial performance with an accounting-based variable. A potential flaw here is the vulnerability of manipulation, since measures like return on assets (ROA) and return on equity (ROE) are established on historical data and companies could be enticed to alter the underlying values by choosing depreciation methods that favour their tax burden (Wang et al., 2014; Wernerfelt and Montgomery, 1988). The return on equity can be seen as a proxy for financial performance whereas the return on assets may estimate the operating performance, according to Alvarez (2012). Different studies combine these two opposing sides of both accounting- and market-based proxies for financial performance (see, e.g. Nakao et al., 2007; Iwata and Okada, 2011). In addition to these financial performance measures that focus on value, two variables introduce measures for risk. McGuire et al. (1988) point out that a lack of social responsibility may increase an individual firm’s financial risk. This is based on the reasoning that investors perceive less socially engaged firms to be riskier investments since management skills are understood to be subpar. Further, McGuire et al. (1988) explain that investors anticipate a rise in firm costs that can be attributed to low investment in social responsibility measures. In contrast, McGuire et al. (1988) state that high degrees of CSR allow firms to reap benefits from more stable relations with the government and the financial community. With reduced market- and accounting-based risk resulting from higher CSR engagement, firms are less sensitive to external shocks. This view is supported by Spicer (1978) who claims that firms rated high on social performance indices have lower total and systematic risk in comparison to the lower-ranked firms. This study includes a measure for the firm’s systematic risk with beta as well as business risk, which is approximated with a firm’s Altman’s Z-score.
if a company is considered an MNC is from 0% to 100%. In coherence with Park et al. (2013) any company that exceeds the threshold of 50% foreign sales to total sales is considered an MNC and receives a dummy value of one. This threshold can be motivated by the fact that when a firm has more than half of its total sales in foreign countries, it can confidently be declared an MNC. All other companies below this threshold of 50% are deemed DCs and receive a value of zero. According to Park et al. (2013), robustness of the results can be increased by using a cut-off point of 20% foreign sales to total sales. This increases the amount of companies that are considered MNCs. Due to a greater amount of MNCs, it can be checked if the results from the first analysis with 50% as threshold are replicable for a larger range of companies. An additional sensitivity check is conducted with the actual percentage of foreign sales to total sales where no indicator variable is created. Further, in coherence with Park et al. (2013) I also use the ratio of foreign assets to total assets as an additional robustness check, as all previous variables were only based on sales figures.
statistics for all variables. A detailed description of all variables, including units of measurement, can be found in appendix A.1.
Table 4: Descriptive statistics of all variables.
Mean Median Min. Max. SD Kurtosis Skewness TQ 0.854 0.614 0.019 4.711 0.836 2.159 8.639 ROA 0.725 0.687 0.024 2.476 0.486 0.891 4.213 ROE 2.417 1.679 -5.755 21.344 3.074 3.334 19.951 BETA 1.068 1.027 0.151 2.590 0.480 0.619 3.400 ZSCORE 1.936 1.834 -1.005 6.257 1.361 0.599 3.489 EPCO 0.262 0.234 0.00002 4.718 0.712 4.465 24.712 EPNOX 0.0008 0.00002 1.52e-10 0.017 0.002 4.305 23.965 EPSOX 0.0008 8.47e-06 1.71e-09 0.015 0.002 4.755 27.611 EPWASTE 0.380 0.001 9.15e-07 25.13 2.753 8.456 74.805 EPENERGY 1.769 0.202 2.71e-10 0.049 4.062 3.558 16.723 EPWATER 0.001 2.46e-06 0.00002 24.347 0.006 6.237 43.427 GROWTH 4.095 3.330 -35.130 61.720 14.337 0.76 6.141 SIZE 18.059 17.614 13.429 24.635 2.538 0.429 2.494 LEVERAGE 0.262 0.252 0.0008 0.669 0.152 0.37 2.704 CAPIN 0.963 0.048 0.001 0.908 0.139 3.574 18.099 MNC_D50 0.463 0 0 1 0.498 0.147 1.021 MNC_D20 0.727 1 0 1 0.445 -1.021 2.042 MNC_C 53.806 54.01 1.42 100 28.667 -0.072 1.870 FA_TA 18.833 6.86 0 93.66 24.683 1.348 3.843
It is notable that the carbon dioxide variable’s mean far exceeds the means of nitrogen and sulphur oxide. This is explained by the fact that companies in general emit more CO2 than
financial performance proxies, a beta of one generally signals that the shares of a company have the same volatility as the market. If beta is smaller than one, the shares are less volatile than the market, whereas a beta greater than one indicates higher volatility than the market. As the mean and median of beta in this study are close to one, it can be assumed that the broad majority of firms have a comparable risk to the market. Altman’s Z-Score is used in financial context to estimate the likelihood that a company may become bankrupt in the future. An Altman’s Z-Score below 1.8 may be understood as an indication for a possible bankruptcy in the future. On the other side, an Altman’s Z-Score close to three can be seen as a good estimation for investors to purchase stock of a particular company. With a mean close to two and a median above 1.8, it can be inferred that most companies in this sample are below the threshold that might signal upcoming financial distress.
4. Results
At first, the results of the Granger causality tests are reported. Subsequently, the results of the multivariate regression analyses are displayed. Then, the results of the regressions with the MNC interaction term are shown and analysed.
4.1 Granger-causality test results
Table 5: Granger causality test with one lag, the table reports the probability values of rejecting the null hypothesis which states that there is no Granger-causality between the variables.
CO2 NOx SOx Waste Energy Water
Unit Tons/ US$ Tons/ US$ Tons/ US$ Tons/ US$ kWh/ US$ Tons/ US$ Number of observations 8150 2630 2520 4700 5770 1300 Tobin's Q does not cause
environmental performance
0.00 0.01 0.00 0.00 0.00 0.00
Environmental performance does not cause Tobin's Q
0.00 0.00 0.00 0.00 0.00 0.00
Return on Assets does not cause environmental performance
0.00 0.00 0.00 0.00 0.00 0.00
Environmental performance does not cause Return on Assets
0.00 0.00 0.00 0.00 0.00 0.00
Return on Equity does not cause environmental performance
0.00 0.00 0.00 0.00 0.00 0.00
Environmental performance does not cause Return on Equity
0.00 0.00 0.00 0.00 0.00 0.00
Beta does not cause environmental performance
0.00 0.00 0.00 0.00 0.00 0.00
Environmental performance does not cause Beta
0.00 0.00 0.00 0.00 0.00 0.01
Z-Score does not cause environmental performance
0.00 0.00 0.00 0.00 0.00 0.00
Environmental performance does not cause Z-Score
0.00 0.00 0.00 0.00 0.00 0.00
has been involved in, Granger-cause financial returns, however the hypothesis of non-Granger causality cannot be rejected for the opposite direction. Due to the difference in the definition of the environmental performance variable, it is difficult to draw a valid comparison between this study and Scholtens (2008). However, it is clear that in Scholtens (2008) environmental concerns Granger-cause financial return and financial risk at a 10% level, which is in line with the results of this study, as concerns can be generally translated into a higher environmental performance due to higher emissions.
Since both the results of table 2 and appendix B.1 do not give clear proof of a unidirectional Granger-causation between respective variables of interest, for the next step the multivariate regression analysis focuses on the negative impact on financial performance from an increase in environmental performance.
4.2 Multivariate regression results
In this section, the results for the multivariate regression analysis are presented. The corresponding hypothesis 2a formulated in chapter 2 states that an increase in emissions and therefore in environmental performance negatively affects financial performance. Table 6 shows the results for all environmental and financial performance proxies.
Table 6: Results for the regression analysis where different environmental performance proxies are regressed on financial performance proxies.
TQ ROA ROE BETA Z-SCORE
The first column of table 6 shows the regression coefficients for the analysis of various environmental performance variables on Tobin’s Q. In appendix D.1, the results for the effect of environmental performance of different pollutants on Tobin’s Q are shown in detail, while using fixed effects estimation. Both, table 6 and appendix D.1 enable inferences about the underlying relationship between environmental and financial performance and display distinctions between several measures for environmental performance. As visible in table 6, the environmental performance variables of energy usage and water emissions have negative coefficients. The regression coefficient for water emissions is evidently larger than the coefficient for energy consumption. Both environmental performance measures for water emissions and energy usage are significant at the 5% level. For all other environmental performance proxies, no significance is present, hence a subsequent analysis of these is not valid. All control variable coefficients are highly significant, most of them at 1% but all at least at 10%. An increase in energy consumption or water emissions thus means that possible investors and market participants understand this as a managerial inefficiency in complying with stakeholder expectations. In consequence the company may be valued lower, which can then be seen through a decrease in Tobin’s Q. The goodness-of-fit, indicated by the R2, is at a
The second column of table 6 shows the regression coefficients when ROA is used as the dependent variable. Results are presented in detail in appendix D.2. There are differences compared to the analysis when Tobin’s Q is selected as the dependent variable, in that three environmental performance measures now have significant coefficients. Here, the results show that the coefficients for CO2 emissions, NOx emissions and energy consumption are negative
when regressed on ROA. Particularly the impact of NOx emissions on ROA is salient, with a coefficient of -21.57. Both coefficients for CO2 emissions and NOx emissions were not
significant in the preceding regression, whereas now they are both significant at the 5% and 1% level, respectively. The coefficient for energy use is again significant at a 1% level, similar to the previous analysis, which used Tobin’s Q as the dependent variable. All control variables have significance at a 1% level. These results mean that for every 1% increase in the three significant environmental performance measures ROA is reduced by the magnitude of the respective coefficient. An increase in emissions, and consequently in the environmental performance variable materializes as reduced net income, which then leads to a reduction of ROA. This effect is strongest for NOx emissions, where a 1% increase in environmental performance leads to a 21.57% decrease in ROA. The R2 statistic is comparable to the prior
analysis, where Tobin’s Q was used as the dependent variable. The outcome of the regression analysis for the three statistically significant variables is in line with hypothesis 2a, as all variables have a negative significant coefficient. For all insignificant coefficients, the null hypothesis cannot be rejected. The findings are in line with the literature. Iwata and Okada (2011) find that greenhouse gas emissions negatively affect ROA. Alvarez (2012) uses the variation in CO2 emissions as a proxy for environmental performance and finds that it
should result in increased net income and thus in higher financial performance. This result can be found in Russo and Fouts (1997) and is consequently comparable to this study, as increases in emissions for this research lead to a decreased return on assets due to managerial inefficiencies. Next, focus is laid on the regression while using another financial accounting-based measure, return on equity, as the dependent variable.
The third column of table 6 depicts the results for the multivariate regression while using ROE as the dependent variable as a proxy for a firm’s financial performance. Appendix D.3 illustrates the results in detail. Similar to the results for ROA in appendix D.2, the environmental performance variable for NOx emissions has a strongly negative coefficient for its regression on ROE. Further, the coefficient for SOx emissions is also negative. The outcome of the regression shows that the emission of nitrogen oxide has an even stronger negative impact on financial performance as proxied by ROE than on operational performance as measured by ROA in the previous paragraph. The magnitude of the impact of sulphur oxide on ROE is also comparatively high, at -24.47. The coefficients for NOx and SOx emissions as proxies for environmental performance are significant at 10% and 5%, respectively. The four remaining coefficients for environmental performance variables have no significance. Most control variables are significant, with only one exception. In an economic sense, a given company that increases its NOx or SOx emissions will experience a reduction in ROE, which can be caused by a decrease in net income. Based on the great negative effect of both coefficients, it can be inferred that a one-percent increase in either emission indicates great violations of stakeholder expectations. It is worth noting that the R2 of all models is low in compared to the models of
and Okada (2011) find a negative impact of greenhouse gas and waste emissions on ROE, however both explanatory variables have no significance in their study. Alvarez (2012), who uses CO2 emission variation in his research also finds negative coefficients on ROE, but these
results are also not significant, which is in line with the results of table 5. Further, the results of this study are in line with Horváthová (2012), as significant negative effects of environmental performance on ROE can be found. Comparing the results of both accounting-based measures, ROA and ROE, with the market-based measure Tobin’s Q, it is apparent that there is no difference in the direction of the effect of an increase in environmental performance on financial performance. Nevertheless, different environmental performance proxies show significance in the regressions for each respective analysis. The next step is to evaluate how an increase in emissions affects the company’s risk measure, which is proxied by beta.
The fourth column of table 6 displays the results when beta is used as a proxy for financial performance. Appendix D.4 shows the results in detail. Four coefficients have positive results. Both NOx and SOx emission coefficients are remarkably higher than CO2 and water
emissions proxies, which is in line with the previous results. An explanation may be that both pollutants are far more detrimental for the environment, hence their excessive emission is punished more severely by a great increase in beta. The coefficients for CO2, NOx and water
seeks to avoid. Similar to the previous analysis, the R2 is comparatively low. The positive
coefficients in appendix D.4 support the underlying hypothesis as they indicate a negative effect on financial performance due to an increase in risk when emissions rise. For the two insignificant coefficients, the null hypothesis cannot be rejected. When compared to the literature, these results for NOx and SOx coefficients are in contrast to Cornell and Shapiro (1987) who stress that the level of social responsibility, to which emissions can be accounted to, only has a minor impact on a firm’s systematic risk. This is reasoned with the argument that all other firms in a market place are not equally affected by an individual’s changing level of social responsibility. Nevertheless, the results of table 6 are in line with Gonenc and Scholtens (2017), who find that an increase in environmental performance also intensifies the systematic risk. However, Gonenc and Scholtens (2017) find a coefficient that is more in line with Cornell and Shapiro’s (1987) hypothesis than the results of this study. Differences in such comparisons may arise due to different focuses on specific industries (see, e.g. Gonenc and Scholtens, 2017) or analyses of individual countries, whereas this study uses a rich international sample across multiple industries. Subsequently, another risk measure, Altman’s Z-Score is utilized to investigate if these effects also occur when another proxy for risk is used as the dependent variable.
The fifth column of table 6 presents the results for the analysis where Altman’s Z-Score is used as the dependent variable. Appendix D.5 shows the regression results in detail. The emissions of CO2 and the energy consumption have negative coefficients, and both are
bankruptcy will increase by the rise in pollution. This is visible for the two significant coefficients for CO2 emissions and energy consumption. The goodness-of-fit, shown by the R2
is at a comparable level to the first two analyses, which enables confidence in the results. The results for regression analysis for CO2 emissions and energy consumption are in line with
hypothesis 2a, while all other environmental performance variables are insignificant, hence the respective null hypothesis cannot be rejected. To the best of my knowledge, there are no studies that have evaluated a possible impact of environmental performance proxies on Altman’s Z-Score. Consequently, I cannot compare these results to other empirical studies that have covered such analysis.
In summary, the previously displayed results have presented evidence that each environmental performance variable has differing effects on the several financial performance variables. The three pollutants carbon dioxide, nitrogen oxide and sulphur oxide can be grouped together as air pollutants. An increase in CO2 emission has a negative effect on ROA and an
As companies from three countries, the United States, United Kingdom and Japan, comprise more than 50% of the entire sample, I analyse the impact of environmental performance on financial performance separate from the full sample. First, a one-way ANOVA test is conducted to investigate differences between the subsamples. The respective outcome is reported in appendix B.4. Since the null hypothesis for this analysis is that there is no statistical difference in the three country’s means, it has to be rejected for the broad majority of analysed variables. Hence, it is assumed that there is variation between the means of the three countries for many variables. The regression results for this sensitivity check are visualized in table 7.
First, it is important to highlight that all respective coefficients for CO2 emissions on
Table 7: Results for sensitivity analysis with overrepresented countries in sample, with all respective environmental performance and financial performance variables.
TQ ROA ROE BETA Z-SCORE
CO2 UK -0.0484*** -0.010 -0.217 0.333*** 0.253* US -0.046*** -0.039*** -0.404*** -0.068*** -0.048** JP -3.665*** -13.477*** -42.303*** -7.581*** 3.288 NOx UK -18.989*** -30.721*** -255.432*** 24.692*** 66.585*** US 9.520 -16.128*** -258.552*** 43.947*** 39.042* JP -839.436*** 1408.578*** 5210.32*** 4090.826*** 2191.622*** SOx UK -35.889*** -51.215*** -442.014*** 43.416*** 51.196* US -3.052 -8.502*** -119.662*** 3.119 -2.128 JP -1434.758*** 2675.446*** 11045.6*** 5832.826*** -2992.122*** WASTE UK -0.027 -0.009 -0.177*** 0.088*** -0.167*** US -0.213*** -0.068*** -0.293 0.137*** -0.084 JP -18.623*** -41.301*** -221.509*** -30.110*** -56.723** ENERGY UK -0.034*** 0.005** -0.021 0.038*** 0.031** US -0.022*** -0.007*** 0.030 0.018*** 0.002 JP -0.402*** 0.459*** 1.416** 0.193 -0.592** WATER UK -54.475*** -24.532 -436.689*** 46.379*** -23.113 US -20.755* -0.308 1.697 61.504*** 65.354*** JP -1224.777 2455.063 1990.053*** 517.796** 552.020 *** p<0.01, ** p<0.05, * p<0.1
increase of emissions occurs, the subsequent reaction from stakeholders in Japan can be more salient. Studies that have analysed the relationship between environmental and financial performance in Japan (see, e.g. Iwata and Okada, 2011; Nishitani and Kokubu, 2012) find that environmentally conscious behaviour is either rewarded financially by stakeholders or non-compliance is punished. Those findings are generally not in line with the results displayed in table 7. Yet, in around 35% of all cases, the direction of the coefficient for the Japanese sample does not match with the full sample displayed in previous tables. Noteworthy are the negative coefficients for CO2 emissions and waste emissions that the Japanese sample shows for the
regressions with beta as the dependent variable. This is in clear contradiction to previous results and the hypothesis as this means that an increase in pollution levels for either variable results in a decrease of systematic risk for Japanese firms.
4.3 Regression results for the interaction effect of multinationality and environmental performance on financial performance
Table 8: Results for regression analysis where a multinationality dummy is interacted with the respective environmental performance proxy and then regressed on each financial performance variable.
TQ ROA ROE BETA Z-SCORE
EPCO -0.119*** -0.082*** -0.235*** -0.020*** -0.154*** MNC_D50 x EPCO -0.153*** -0.009 -0.154*** 0.076*** 0.128*** EPNOX -32.97*** -15.21*** -83.22*** 18.79*** 21.86** MNC_D50 x EPNOX -12.06** 5.038 14.14 -9.766 16.14 EPSOX -21.18*** -14.35*** -86.91*** -0.737 -17.33*** MNC_D50 x EPSOX -11.46 6.512 -34.03 14.23 -36.21 EPWASTE -0.024*** -0.016*** -0.001 -0.002 -0.089*** MNC_D50 x EPWASTE 0.012** 0.009*** -0.020** 0.007 0.064*** EPENERGY -0.033*** -0.011*** 0.013 0.024*** -0.027*** MNC_D50 x EPENERGY 0.0007 0.002* -0.055** -0.014*** 0.025*** EPWATER -11.00*** -0.183 15.50* 9.963*** 3.383 MNC_D50 x EPWATER 12.06*** -1.367 -47.13*** -11.96*** 4.580
The first column of table 8 visualizes the outcome when Tobin’s Q is used as the financial performance proxy. The results are presented in detail in appendix D.6. All coefficients of the environmental performance variables have a negative direction when regressed on Tobin’s Q. The interaction terms for CO2 emissions and NOx emissions are both
The negative and significant interaction terms materialize in that the relationship between higher emissions of CO2 and NOx and reduced Tobin’s Q is weaker for MNCs than for DCs.
Hence, they experience decreased financial losses arising through higher emission levels. The opposite is true for waste and water emissions, where a negative impact of the relationship between higher emissions on Tobin’s Q is strengthened for the case of multinational corporations. The R2 of the models is at a comparative level to previous regression analyses in
this paper. The results for waste and water emissions are in line with hypothesis 2b, as the interaction term is positive and significant, which translates to a stronger negative effect of increased emissions on financial performance for MNCs in comparison to DCs. The null hypothesis of hypothesis 2b cannot be rejected for the interaction terms of SOx emissions and energy consumption. As the interaction terms for CO2 and NOx are significant yet negative, it
can be concluded that they are not in line with hypothesis 2b. Next, the same methodology is applied to regressions where ROA is used as the dependent variable to investigate if a similar pattern of results can be found for an accounting-based dependent variable.
The second column of table 8 supports the previously found results for the effect of the coefficients from environmental performance on ROA. Detailed results are shown in appendix D.7. The regression coefficients for CO2, NOx, SOx and waste emissions as well as energy
experience greater losses in net income by an increase in the environmental proxy compared to their domestic counterparts. The R2 for all models in table 10 are higher than those of the
comparable models in this study, hence indicating superior explanatory power. Both of these interaction terms are in line with hypotheses 2b. The null hypothesis cannot be rejected for the remaining interaction terms due to lack of significance of the coefficients. The next step is to analyse if similar results appear when another accounting-based measure for financial performance, ROE, is used as the dependent variable.
The third column of table 8 shows the results for testing hypothesis 2b with ROE as the dependent variable. Appendix D.8 presents the outcome of the analysis in depth. The regression coefficients for the environmental proxies of CO2, NOx and SOx emissions are all negative.
The magnitude of the latter two is even greater than those displayed in table 6, where the regression was conducted without an interaction term. Further, the interaction terms for carbon dioxide, waste and water emissions as well as energy consumption are negative, with the interaction coefficient for water emissions being salient in its great magnitude. The three regression coefficients for the environmental proxies regressed on ROE are significant at 1%. Both interaction terms for CO2 and water emissions are significant at 1%, whereas the
the previous two dependent variables. For the three environmental performance proxies for carbon dioxide, nitrogen oxide and sulphur oxide emissions, the coefficient is in line with hypothesis 2a. However, the interaction term for carbon dioxide emissions contradicts hypothesis 2b as it is negative. The regression coefficient as well as the interaction term for water emission are not in line with each respective hypothesis. For all insignificant coefficients and interaction terms, the corresponding null hypothesis cannot be rejected. In the subsequent analysis, beta as a risk proxy is used to investigate if the interaction terms are in line with hypothesis 2b for this dependent variable.
In the fourth column of table 8, outcomes for the regression with MNC interaction are shown for beta as the dependent variable. Appendix D.9 presents a detailed table for these regressions. The regression coefficient of CO2 emissions is negative, whereas the coefficient
for NOx and water emissions as well as energy consumption are positive. Furthermore, the interaction term for carbon dioxide emissions is positive while for energy consumption and water emissions it is negative. The regression coefficients of carbon dioxide, sulphur oxide, water and energy are significant at 1%. Moreover, the interaction terms for carbon dioxide, energy and water are significant at 1%. Most control variables are highly significant at 1%, only the firm size does not show consistent significance for different types of environmental performance proxies. The positive interaction term for carbon dioxide emissions means that the increase in such emissions leads to a decrease in beta and this effect is stronger for MNCs. The opposite holds true for energy consumption and water emission, where an increase in either variable leads to an increase in risk. However, both interaction terms for energy consumption and water emission show that such an effect is weaker for MNCs in comparison to DCs. Hence, MNCs do not experience an increase in their risk beta as great as DCs do for these two variables. All regressions in appendix D.9 have slightly lower explanatory power, with lower R2 compared
to previous models. The coefficient for CO2 emissions regressed on beta is not in line with
2b in that MNCs experience a stronger effect of the underlying relationship than DCs. The regression coefficients for NOx and water emissions as well as energy consumption are in line with hypothesis 2a. For the latter two, the interaction term however cannot confirm hypothesis 2b. The null hypothesis for either hypothesis 2a or 2b cannot be rejected for all insignificant results. As a next step, Altman’s Z-Score, which is another proxy for the firm’s risk, is used as the dependent variable to analyse if comparable results are present.
The final column of table 8 shows the result of the analysis when Altman’s Z-Score is used as the dependent variable. Detailed results can be found in appendix D.10. The regression coefficients of the environmental proxies for carbon dioxide, sulphur oxide, waste and energy are negative while only the coefficient for nitrogen oxide is positive. Positive interaction terms can be found for carbon dioxide, waste and energy. All four negative regression coefficients are significant at 1% whereas the positive coefficient for nitrogen oxide is significant at 5%. The three positive interaction terms are all significant at a 1% level. With a few exceptions, all control variables are significant, mostly even at 1%. The three coefficients for carbon dioxide, waste and energy indicate that an increase in emissions leads to a decrease in the Z-Score, which brings the company closer to a possible threat of bankruptcy. This effect is stronger for MNCs, as indicated by the significant and positive interaction term. This means that multinational corporations face more severe implications for their Z-Score than domestic corporations when either emission level increases. The explanatory power of these analyses approximated by R2
is comparable to the analyses when Tobin’s Q has been used as the dependent variable. All significant regression coefficients are in line with hypothesis 2a except for the nitrogen oxide coefficient, which is positive. The three significant interaction terms are in line with hypothesis 2b since they are positive. For all insignificant coefficients and interaction terms, the respective null hypothesis cannot be rejected.
results are independent from using sales measures as indicator. Results for this analysis can be found in appendix E.11 through E.15. These regression outcomes are generally in line with the results found in this chapter as more than half of the results are similar in direction and significance. Therefore, it can be argued that the results found in table 8 are robust even when the indicator is changed from sales to asset ratios. The regression coefficients as well as the coefficients for the interaction terms are least in coherence with the results for the respective regression when ROA is used as a dependent variable, since five coefficients are not in the same direction as the previously shown results. To conclude, it is visible that all three robustness checks for this analysis generally produce similar results and thus confirm the sensibility of the conducted research on the interaction between multinational corporations and environmental performance.
5. Conclusion
This study analyses if MNCs experience stronger negative effects on financial performance from an increase in environmental performance in comparison to domestic companies. The statistical results show partial evidence that MNCs suffer financially more than domestic corporations from an increase in emissions and a subsequent increase in environmental performance. A stronger negative effect on financial performance from an increase in emissions can partially be found for CO2,waste and water emissions as well as
relationship between environmental and financial performance. A sample of 819 firms from 38 countries is used to enable inferences. A possible limitation of this study is the inclusion of all available countries and industries in the sample in order to construct a large sample. Even though fixed effects regressions account for country- and industry-specific variation, a full explanation of differences may not be possible. Research that builds upon this study could investigate both country- and industry-specific results in more depth. In addition, the sample may be biased for the evaluation of a stronger negative effect from higher environmental performance on financial performance for MNCs. This bias may occur as it can be expected that certain developed countries are overrepresented, as they simply have more MNCs than other nations.
Five different proxies for financial performance are used in this study, with Tobin’s Q as a market-oriented measure, return on assets and return on equity as accounting-based proxies and beta and Altman’s Z-Score as estimates for risk. In order to conceptualize environmental performance, six different emission types that pose a threat to the environment are selected. Three of these, CO2, NOx and SOx emissions, address air pollution. Further, proxies for waste
and water pollution are used in addition to the energy consumption by a company. To measure the interaction effect for the main research question, the foreign sales to total sales ratio of a company is utilized and a firm is deemed a multinational corporation when the threshold of 50% is exceeded. For employing a robustness check, three other measures for determining an MNC are employed. All quantitative information for these variables is collected from Thomson Reuters Datastream and ASSET4 ESG database.
impact of increased emissions proxied by higher environmental performance on financial performance. This paper stresses that the impact of emissions on financial performance differs for each respective proxy, which can be seen as confirmation that investors value various emissions differently. These differences in the negative effect for every emission proxy occur because the scope of the environmental issue each pollutant causes is different. These issues can either be local or global, and the length of time until damages emerge varies (Iwata and Okada, 2011). Further, pollutants can also differ in the severity of damages. Based on this, it can be argued that different stakeholders focus on different aspects of environmental issues (Iwata and Okada, 2011). Stakeholders can be affected differently by an increase in pollutant emissions. Local communities may experience more direct effects, whereas other stakeholders are only affected through financial relationships with individual corporations (Iwata and Okada, 2011). Consequently, for specific groups, global warming appears to be the more pressing issue, whereas for other stakeholders, the local effects of an increase in emissions are more detrimental.
another proxy for estimating an effect on financial performance. Toxic waste is not included in this study due to data constraints. Water emissions negatively affect the market-based measure Tobin’s Q and increase the firm’s risk, proxied by beta. When a company increases its energy consumption, which is the sixth environmental performance proxy, negative effects result for Tobin’s Q, an accounting-based measure in ROA and the company’s bankruptcy risk, proxied by Altman’s Z-Score.
As three countries are overrepresented in this sample, I separately test for their individual results to ensure robustness of the main regression outcomes. While the results for the United States and the United Kingdom are broadly in line with the general sample results, the outcome of the regression analysis of Japanese companies presents a different picture. Various regression coefficients for the respective emission proxies are substantially greater than those of the other two countries compared in this sub-sample. Moreover, the coefficient is often in the opposite direction of the two other countries and the entire sample while still presenting highly significant results. This may be due to dissimilarities in reporting standards and legislation that influence the behaviour of a corporation towards emission of harmful pollutants. Including such factors in future research may enhance the results found in this study. Further, it could be hypothesized that based on stricter laws, investors and other stakeholders in Japan are considerably more sensitive to increases in emissions. In line with the stakeholder theory, Japanese investors and stakeholders consequently punish the corporations more drastically for their mistakes.
The results produced from the regressions including the interaction effect partially indicate that MNCs experience a stronger negative effect on financial performance than domestic companies when there is an increase in emissions. Evidence for this stronger negative effect can be found for CO2 emissions only when its impact on Altman’s Z-Score is evaluated.
Appendices
Table A.1: Full description of variables employed in this study. The source for all variables is Thomson Reuters Datastream.
Name Unit Calculation
Tobin’s Q US $ Market value / Total Assets
Return on Assets US $ Net Sales / Total Assets
Return on Equity US $ Net Sales / Shareholder Equity
Beta Regression of firm’s stock market return
against local market index return
Altman’s Z-Score US $ Z = 1.2A + 1.4B + 3.3C + 0.6D + 1.0E
A = Working Capital / Total Assets B = Retained Earnings / Total Assets
C = EBIT / Total Assets D = MV Equity / Total Liabilities
E = Net Sales / Total Assets
CO2 Tonnes / US $ Total CO
2 emissions / Net Sales
NOx Tonnes / US $ Total NOx emissions / Net Sales
SOx Tonnes / US $ Total SOx emissions / Net Sales
Waste Tonnes / US $ Total Waste produced / Net Sales
Energy kWh / US $ Total Energy use / Net Sales
Water Tonnes / US $
Total Water Pollutant emissions / Net Sales
Growth Percent Net Sales annual growth
Size US $ Natural logarithm of Total Assets
Leverage US $ Total Liabilities / Total Assets
Capital Intensity US $ Capital expenditure / Net Sales
MNC_D50 Percent Foreign Sales/ Total Sales dummy at 50%
threshold
MNC_D20 Percent Foreign Sales/ Total Sales dummy at 20%
threshold
MNC_C Percent Foreign Sales/ Total Sales
Table A.2: Overview of all countries represented in this sample together with the respective number of companies for each country
Country No. of companies Country No. of companies
Austria 4 Luxemburg 1
Australia 40 Mexico 3
Belgium 7 Malaysia 1
Brazil 7 Netherlands 18
Canada 28 Norway 10
China 1 New Zealand 4
Colombia 1 Portugal 4
Czech Republic 1 Russia 6
Denmark 12 Sweden 25
Finland 12 Singapore 2
France 49 South Africa 9
Germany 35 South Korea 13
Greece 7 Spain 24
Hong Kong 10 Switzerland 19
Hungary 1 Thailand 2
Israel 1 Taiwan 5
India 5 United Kingdom 118
Italy 17 USA 137
Japan 180
Total 819
Table A.3: Overview of the SIC-industries represented in the sample with the respective number of companies for each SIC-industry
Industry SIC - Code No. of companies
Agriculture, Forestry, Fishing 0-999 2
Mining 1000-1499 56
Construction 1500-1699 21
Manufacturing 2000-3999 398
Transportation, Communication, Electric, Gas and Sanitary services
4000-4999 146
Wholesale Trade 5000-5199 8
Retail Trade 5200-5999 32
Finance, Insurance and Real Estate 6000-6799 116
Services 7000-8999 40
Table B.1: Breusch-Pagan/ Cook-Weisberg test for heteroscedasticity for a linear regression model for this sample, the results show the probability that the null hypothesis can be rejected. The corresponding null hypothesis states that the error variances are all equal. The alternative hypothesis states that the error variances increase or decrease as the predicted value of the independent variables, in this case environmental performance proxies, increase.
Prob. > 𝜒- Tobin’s Q ROA ROE BETA Z-SCORE
EPCO 0.000 0.000 0.000 0.141 0.000 EPNOX 0.000 0.000 0.000 0.018 0.000 EPSOX 0.000 0.000 0.000 0.031 0.000 EPWASTE 0.000 0.000 0.000 0.000 0.000 EPENERGY 0.000 0.000 0.000 0.459 0.000 EPWATER 0.000 0.000 0.000 0.000 0.000
Table B.2: Wooldridge test for autocorrelation in panel data, the results show the probability that the null hypothesis can be rejected. The respective null hypothesis is that there is no first-order autocorrelation, which means that the errors are uncorrelated with each other.
Prob. > F Tobin’s Q ROA ROE BETA Z_SCORE
EPCO 0.000 0.000 0.000 0.000 0.000 EPNOX 0.000 0.000 0.205 0.000 0.000 EPSOX 0.000 0.000 0.528 0.000 0.000 EPWASTE 0.000 0.000 0.003 0.000 0.000 EPENERGY 0.000 0.000 0.000 0.000 0.000 EPWATER 0.000 0.000 0.396 0.000 0.000
Table B.3: Correlation matrix for all variables used in this study
TQ ROA ROE BETA Z_SCORE
LEV -0.334 -0.039 0.302 0.140 -0.356 CAPIN 0.059 -0.349 -0.250 0.070 -0.020 MNC_C 0.252 -0.122 -0.074 0.003 0.173 MNC_D50 0.177 -0.080 -0.044 0.039 0.134 MNC_D20 0.042 -0.055 -0.015 0.057 0.005 FATA 0.063 -0.005 -0.044 0.004 0.087
EPCO EPNOX EPSOX EPWASTE EPENERGY EPWATER
EPCO 1 EPNOX 0.501 1 EPSOX 0.471 0.917 1 EPWASTE 0.098 0.013 0.098 1 EPENERGY 0.732 0.433 0.354 0.136 1 EPWATER 0.194 0.396 0.372 0.005 0.367 1 GRO 0.019 0.023 0.027 -0.014 0.020 0.032 SIZE -0.431 -0.202 -0.177 -0.154 -0.459 -0.260 LEV 0.073 0.037 0.031 -0.096 0.035 0.031 CAPIN 0.355 0.306 0.229 0.160 0.303 0.045 MNC_C 0.165 0.015 0.008 0.076 0.247 0.094 MNC_D50 0.134 -0.064 -0.050 0.087 0.165 0.019 MNC_D20 0.051 0.046 0.047 0.028 0.096 0.031 FATA 0.143 -0.061 -0.070 0.140 0.245 0.120
GRO SIZE LEV CAPIN
